Google's 'Woke' AI Troubles: Charting a Pragmatic Course

 

As Google CEO Sundar Pichai informed employees in a note on Tuesday, he is working to fix the AI tool Gemini that was implemented last year. The note stated that some of the text and image responses reported by the model were "biased" and "completely unacceptable". 

Following inaccuracies found in some historical depictions generated by its application, the company was forced to suspend its use of its tool for creating images of people last week. After being hammered for almost a week last week over supposedly coming out with a chatbot that could be used at work, Google finally apologised for missing the mark and apologized for getting it wrong. 

Despite the momentum of the criticism, the focus is shifting: This week, the barbs were aimed at Google for what appeared to be a reluctance to generate images of white people via its Gemini chatbot, when it came to images of white people. It appears that Gemini's text responses have been subjected to a similar criticism. 

In recent years, Google's artificial intelligence (AI) tool Gemini has been subjected to intense criticism and scrutiny, especially as a result of ongoing cultural clashes between those of left-leaning and right-leaning perspectives. In contrast to the viral chatbot ChatGPT, Gemini has faced significant backlash as a Google counterpart, demonstrating the difficulties associated with navigating AI biases. 

As a result of the controversy surrounding Gemini, images that depict historical figures inaccurately were generated, and responses to text prompts that were deemed overly politically correct or absurd by some users, escalated the controversy. It was quickly acknowledged by Google that the tool had been "missing the mark" and the tool was halted. 

However, the fallout from the incident continued as Gemini's decisions continued to fuel controversies. There has been a sense of disempowerment among Googlers on the ethical AI team during the past year, as the company increased the pace at which it rolled out AI products to keep up with its rivals, such as OpenAI, who have been rolling out AI products at a record pace. 

Gemini images included people of colour as a demonstration that the company was considering diversity, but it was also clear that the company failed to take into account all possible scenarios in which users might wish to create images. 

In her view, Margaret Mitchell, former co-head of Google's Ethical AI research group and chief ethics scientist for Hugging Face AI, has done a wonderful job of understanding the ethical challenges faced by users. As a company that had just been established four years ago, Google had been paying lip service to increasing its awareness of skin tone diversity, but it has made great strides since then.

As Mitchell put it, it is kind of like taking two steps forward and taking one step backwards." he said. There should be recognition given to them for taking the time to pay attention to this stuff. In a general opinion, Google employees should be concerned that the social media pile-on will make it even harder for internal teams who are responsible for mitigating the real-world harms that their artificial intelligence products are causing, such as whether the technology can hide systemic prejudices. 

The employees worry that Google employees should not be able to accomplish this task by themselves. A Google employee said that the outrage that was generated by the AI tool for unintentionally sidelining a group that is already overrepresented in the majority of training datasets could spur some at Google to argue for fewer protections or guardrails on the AI’s outputs — something that, if taken to an extreme, could hurt society in the end. 

The search engine giant is currently focused on damage control as a means to mitigate the damage. It was reported that Demis Hassabis, the director of Google DeepMind's research division, said on Feb. 26 that the company plans to bring the Gemini feature back online within the next few weeks. 

However, over the weekend, conservative personalities continued their attack against Google, specifically in light of the text responses Gemini provides to users. There is no doubt that Google is leading the AI race on paper, with a considerable lead. 

The company makes and supplies its artificial intelligence chips, has its cloud network, which is one of the requisites for AI computation, can access enormous amounts of data, and has an enormous base of customers. Google recruits top-tier AI talent, and its work in artificial intelligence enjoys widespread acclaim. A senior executive from a competing technology giant expressed to me the sentiment that witnessing the missteps of Gemini feels akin to observing a defeat taken from the brink of victory.